---
title: "Data Hub: overview"
url: "https://ensights.productfruits.help/en/article/data-hub-overview"
language: "en"
last_updated: "2026-08-26T13:05:53.800Z"
---

# Data Hub: overview

How data gets into enSights and back out again: connectors, the API, MCP, and how complete your setup data is.

## What Data Hub is for

Every number in enSights arrived from somewhere — an inverter portal, a meter, a SCADA system — and some of it needs to go somewhere else afterwards. Data Hub is where those routes are managed.

Two of its pages bring data *in*, two take it *out*, and one tells you whether what came in is complete enough to be useful:

![](https://cdn-assets.productfruits.com/3925ae9c-39da-4dd8-8bf9-7991b9d10de3)

## The four pages

| Page | Direction | What it does |
| --- | --- | --- |
| APIs | Out | API keys your systems use to read enSights data, scoped to particular assets, plus usage statistics per key. |
| MCP | Out | The Model Context Protocol endpoint, which lets AI tools such as Claude, ChatGPT or Cursor query your enSights data directly. |
| Connectors | In | Credentialed connections to monitoring vendors — SolarEdge, SMA, Enphase, Sungrow and around forty others — that pull production data into the platform. |
| Data Onboarding | Quality | How completely each system has been set up, scored field by field, so you can see what is missing before it distorts a report. |

## How they fit together

The chain runs in one direction:

1.  A **Connector** authenticates to a vendor portal and pulls in readings.
2.  Those readings become the production, alert and performance figures you see across the platform.
3.  **Data Onboarding** scores how much of the surrounding context — capacity, location, expected production, financials — has actually been filled in. Readings without that context can be displayed but not judged.
4.  **APIs** and **MCP** hand the result back out: the API to your own software, MCP to AI tools.

That order matters when something looks wrong. A missing figure is a connector problem; a figure that exists but cannot be compared to anything is an onboarding problem; a figure your own system cannot retrieve is an API scoping problem.

## Two things to know before you start

### This page holds credentials

Connectors store vendor usernames and passwords, and the APIs page stores keys and secrets, because the platform has to act on your behalf on a schedule. Anyone who can open these pages can read those values. Restrict access accordingly — the relevant permission in Roles & Permissions is **Datapoints**.

### Poll frequency has a cost

Vendor APIs are rate-limited, and most contracts cap how many calls you may make. Connectors therefore let you set the resolution at which each timeframe is fetched, and include an **API Call Estimator** so you can see the monthly call volume a configuration implies before you commit to it. Setting everything to the finest resolution across a large fleet is the usual way an account runs into a vendor limit.

## Where to go next

-   **Adding a new site's data feed?** Connectors.
-   **A site's data stopped arriving?** Connectors first, to check the connector's status, then Settings → Activity → Processes for the failed import.
-   **Building an integration?** APIs, and the linked API documentation.
-   **Want to ask questions of your fleet from an AI tool?** MCP.
-   **A report has gaps, or performance figures look impossible?** Data Onboarding — it is usually a missing field rather than a missing reading.